The Journal of Neuroscience
June 1966, 6(S): 1781-1795
Pioneer Growth Cone Steering Along a Series of Neuronal and
Non-Neuronal Cues of Different Affinities
Michael Gaudy* and David Bentleyj*Biophysics Group and TDepartment of Zoology, University of California, Berkeley, California 94720
We have analyzed the morphology of over 5000 Til pioneer
growth cones labeled with anti-HRP, which reveals the disposition of axons, growth cone branches, and filopodia. Til axon
pathways typically consist of a sequence of 7 characteristically
oriented segments, with a single, distinct reorientation point between each segment. Growth cones exhibit the same orientations and reorientations in a given region as do axon segments
at later stages. The single, distinct reorientations suggest that
growth cones make discrete switches between guidance cues as
they grow.
Til growth cones are guided by various types of cues. A set
of 3 immature identified neurons serves as nonadjacent guidepost cells and lies at the proximal end of 3 of the axon segments.
To form another segment, growth cones reorient along a limb
segment boundary within the epithelium. Growth cones also respond consistently to, and orient toward, a specific mesodermal
cell, which may be a muscle pioneer. Thus, growth cones respond to at least 3 different types of cells in the leg.
Til growth cones exhibit a hierarchy of affinity for these cues.
Guidepost neurons are the dominant cues in that contact with
them reorients growth cones from guidance by the other types
of cues.
Growth cone branches are exclusively oriented to specific cues.
Growth cones reorient by extending a branch directly to the cue
of highest affinity and by withdrawing any branches that are
extended to a cue of lesser affinity. A single filopodium in direct
contact with a guidepost neuron can reorient a growth cone that
still has multiple filopodia or even prominent branches specifically oriented to a previous cue of lesser affinity. These observations suggest that growth cone steering may not result simply
from passive adhesion and filopodial traction, but may involve
more active processes.
(Til) pioneer axons extend proximally from their cell bodies,
and then turn and grow circumferentially before making a second turn and continuing proximally to the CNS (Bentley and
Keshishian, 1982a, b; Ho and Goodman, 1982).
Along their route, the Til axons extend between a set of
putative guidepost neurons, with which they selectively dyecouple. Of these neurons, the Trl neuron and (pair of) Cxl
neurons are located at the characteristic turns on the route (Bentley and Keshishian, 1982a, b; Ho and Goodman, 1982; Keshishian and Bentley, 1983a-c). Selective ablation of the Cxl
guidepost cells has shown them to be necessary for normal pathfinding (Bentley and Gaudy, 1983a, b). Another cue for Til
growth cones is a limb segment boundary, which orients them
circumferentially (Bentley and Caudy, 1983b; Caudy and Bentley, 1986). There also is evidence that a proximal increase in
epithelial affinity within segments can guide Til growth cones
proximally when they are not in contact with higher affinity cues
(Caudy and Bentley, 1986).
Here we describe the detailed route taken by the Til growth
cones in normal embryos, and their characteristic interactions
with identified guidepost neurons and other extrinsic cues. These
interactions, and the detailed dispositions of growth cone
branches and filopodia, suggest that the Til growth cones have
different affinities for their different cues. These interactions and
dispositions also are interpreted with regard to their implications for models of growth cone steering based on adhesionmediated(Bray 1979,1982; Letoumeau, 1975a, b, 1979,1983);
or active (Cooper and Schliwa, 1985) responses to guidance cues.
During the course of development, axons grow along specific
pathways to their appropriate targets. Many axons are guided
by fasciculating with previously formed axons and nerves. However, the initial axons growing through any region must “pioneer” a pathway through an axonless environment, and therefore must be guided by other types of cues. We are investigating
the guidance mechanisms for afferent pioneer axons in the legs
of grasshopper embryos.
Bate (1976) identified the first peripheral axons to arise in
embryonic legs and antennae of grasshoppers as those of pairs
of neurons located at the tips of those appendages. In legs, these
Received Sept. 13, 1985; revised Dec. 6, 1985; accepted Dec. 18, 1985.
We thank Alma Toroian-Raymond for light-microscope histology and technical
assistance, and Dr. Marty Shankland for critiquing the manuscript. Support was
provided by NIH NS09074-14and NIH 2T32-GM07379. D.B. wasalso supported
by an NIH Jacob Javits Award, and M.C. by an Einstein Fellowship for Developmental Neuroscience, from the University of California.
Correspondence should be addressed to Michael Caudy, Department of Physiology, University of California, San Francisco, CA 94143.
Copyright 0 1986 Society for Neuroscience 0270-6474/86/06 178 l-15$02.00/0
1781
Materials and Methods
Antibody
labeling
The anti-HRP labeling protocol is described in detail in Caudy and
Bentley (1986). In brief, it consists of the following steps:(1) fixation
in 4% formaldehyde in physiological saline; (2) 1 hr rinse in PBS with
Triton
(“rinse”);
(3) 12-24 hr incubation
in 0.02% rabbit anti-HRP
serum; (4) 1 hr wash in rinse with 1% BSA; (5) 12-24 hr incubation in
0.04% TRITC-coniueated
goat anti-HRP
(InG fraction): 161 1 hr wash
in rinse with 1% iSi;; anb(7) whole-mod: under co&&s
with 40
pm wire spacersbetween coverglassand slides in 1:9 PBS:glycerol
mountant with antioxidants. Using this labeling technique, we have
examined the morphologies of approximately
5000 Til growth cones
(2 Ti 1s/leg) in approximately 500 animals from about 50 different clutches
of eggs.
Terms
Filopodia
usually can be clearly distinguishedfrom branches.Filopodia
are the smallest diameter processesobserved(~0.5 pm) and typically
are very uniform in diameter from their tip to within 2 or 3 pm of their
base. Filopodial clusters are several filopodia extended in a highly oriented group with few or no filopodia extended within the adjacent
regions. Branches of growth cones are any processes larger than a filopodium or smaller than an axon. They can vary widely in diameterfrom 0.5 to about 5 Km-and also can form branches of their own.
1782
Vol. 6, No. 6, Jun.
Caudy and Bentley
Table 1. Mean length of, and guidance cues for, the characteristic
Til axon segments”
Mean length
@rn + SD)
Segment
Known or putative
cues
a
=
37.7 i 16.1
1. Intrinsic polarity of the Ti 1 cell
bodies
2. Proximal increase in epithelial
b
=
54.5 * 12.1
Fe1 guidepost neuron
c
=
27.6 + 10.1
1. Proximodorsal orienting factor
2. ml mesodermal cell (probable
muscle pioneeti)
3. Other dorsally located cells, e.g.,
Table 2.
1986
Percentage of Til guidance by guidepost neurons between
eachpair of neuronson the path
Neuron pair
Ti 1 guidance (%)
From Ti 1 to Fe1
From Fe1 to Trl
From Trl to Cxls
59
67
66
affinity
m2
4. Femur-trochanter
boundary
segment
d
=
55.7 -t 6.1
Tr 1 guidepost neuron
;
=
38.2
73.7 kf 21.7
19.2
Distal segment boundary
Cx 1 guidepost neurons
g
=
127.6 f 23.8
of coxa
I. Unidentified, strong, proximal
orientation factor
2. p cells
Total = 416.4
a, b: n = 10; c-g: n = 11
u Segments are depicted in Figure 10.
h See text; cell depicted in Figure 5A (Ball et al., 1985).
c Total of mean segment lengths traversed by Ti I growth cones.
Measurementof axon segments
The length of the a segment (Fig. 10) is measured as the distance between
the point where the Ti 1 axons emerge from their cell bodies to the point
where they reorient toward the Fe1 cell. The length of the b segment is
the distance between the reorientation toward the Fe1 cell to the point
of contact with that cell, plus the length of axon running along the Fe1
cell to the point where the axon leaves it. The remaining axon segments
are measured similarly.
Definition of cells,axes,and stagesof development
The names of the various cells and limb axes are defined in the Materials
and Methods section of Gaudy and Bentley (1986), as are the stages of
neuronal differentiation. General staging of embryogenesis is described
in Bentley et al. (1979).
Results
Characteristic Til Axon and Growth
Cone Morphologies
Til axons closelyfollow the initial
growth coneroutes
Ti 1 axons are not displaced laterally, or “dragged” (Gunderson
and Park, 1984), over the epithelial substrate by the motile
growth cone. Distinct axon reorientations are regularly observed
(Figs. 1, B, D, 2, A, B, 3C), indicating that the axons are firmly
attached to the substrate, at least at the point of reorientation.
In some cases, axon reorientations are observed immediately
behind the growth cone (Fig. 3C9, suggesting that the axons are
attached to the substrate either as they are laid down, or very
soon thereafter. In addition, Til axons generally have distinct
edges flattened against the epithelial substrate at all points along
their length, an indication that they are attached all along their
length.
Distinct reorientations occur at characteristic locations (see
below), and are observed at the same location both soon after
the growth cone reorients and at substantially later stages. In
Figure 1, B and D, a reorientation that is frequently observed
about 30-50 ym proximal to the Til cell bodies is present both
when the growth cone is at the Fe1 cell and about 6 hr later
when the growth cone is at the Trl cell. These observations
indicate that the distinct reorientations observed in axon morphologies accurately reflect the location and direction of reorientations made by growth cones at earlier stages.Therefore, the
morphologies of recently established, or “nascent,” axons can
be used to determine the precise routes taken by growth cones
in a particular leg.
Til axonsfollow a stereotypedroute consistingof
distinctly orientedsegments
Observation of growth cone and nascent axon morphologies at
various points along the path has shown that the overall Til
axon morphology consists of distinct, highly stereotyped segments of axon. In normal pathways,
a series of straight, characteristically oriented segments of axon is observed (Figs. 1,
B, D, 2, 3, and 7). In the most complex case of such normal
pathways, the overall path is comprised of 7 distinct axon segments, which are schematized and labeled in Figure 10. The
average lengths of these axon segments are given in Table 1 (see
also Table 2). We will describe the most complex case first;
variants in which certain segments are missing will be discussed
later.
Segment a: The segment of initial outgrowth of the pair of
Til axons from their cell bodies is oriented proximally along
the geometric axis of the tibia (Fig. 1, A, B, D; schematized in
Fig. 10).
Segment b: The next segment is initiated by a single, distinct
reorientation and is oriented in the proximoventral direction
(Fig. 1, B, D). This segment extends straight in this direction to
the Fe 1 identified neuron and then extends along the dorsal edge
of that neuron (Fig. 10). The segment is terminated where the
axons leave the proximal edge of the Fe1 cell and distinctly
reorient (Fig. 1, C, D).
Segment c: On leaving the Fe 1 neuron, Ti 1 axons (Figs. 1D,
2A) and growth cones (Figs. 1, B, C, 5A, 64 again distinctly
reorient, this time in the proximodorsal direction. This reorientation is more acute than is necessary to orient proximally
along the geometric axis of the limb (cJ: drawing in Fig. lo), so
that the c segment always is oriented dorsal to the proximally
located Trl neuron (Fig. 1, C, D).
Segment d: This segment begins with a distinct reorientation
from the c segment and is oriented proximoventrally (Figs. 1D;
2, A, B). The d segment extends directly to the Trl neuron and
then along that neuron to its ventral edge (Figs. 40, 5D).
Segment e: This segment is oriented circumferentially from
its origin at the Trl cell (Figs. 2B, 3, 5D; see also Bentley and
Caudy, 1983b). It extends straight ventrally and, as a result, is
oriented toward a point that lies about 20 pm straight distal to
the pair of Cxl identified neurons (Fig. 3C). The e segment
terminates with a distinct reorientation toward the Cxl guidepost neurons.
Segmentf: This segment is oriented toward the Cxl neurons
The Journal
of Neuroscience
PioneerGrowth Cones Steer to Different Guidance Cues
1783
Figure 1. Til morphologies along the u, b, c, and d segments of the limb. A, The pair of Til neurons has begun proximal axonogenesis, and
their axons are extended straight proximally along the geometric axis of the limb. This forms the a segment (see text). The Til growth cones
typically have no branches on the a segment. The Ti 1 cell bodies also are aligned along the proximodistal axis of the leg. Curved arrow, Cx 1 cells.
Calibration bar, 50 pm. B, The Til cells are aligned along the limb axis and have extended axons proximally along the a segment. The axons
distinctly reorient at a point (upward arrow) about 40 pm proximal to the Til cell bodies and then extend directly to the site (arrowhead) where
the Fe1 cell arises. This forms the b segment (see text). From the Fe1 site, 2 clusters of filopodia (downward arrow) are extended proximodorsally
in the direction of a pair of anti-HRP labeled cells (asterisk) in the epithelium. Curved arrow, Cxl cells. Calibration bar, 50 pm. C, The Til axons
have extended proximally from their cell bodies and have contacted the Fe 1 cell (solid arrowhead). Growth cones have numerous lateral filopodia,
about 25-35 pm in length, and have extended proximodorsally from the Fe1 cell, along the c segment (see text). A cluster of leading filopodia
(downward arrow) is extended directly toward the large, lamellar ml cell (caret), which is labeled with anti-HRP. From the same growth cone, a
branch is extended proximoventrally along a single filopodium (upward arrow) that is in direct contact with a labeled Trl cell (open arrowhead).
Calibration bar, 50 pm. D, The Til axons have extended proximodorsally from the Fe1 cell (solid arrowhead) to a point (upward arrow at left)
from which they have reoriented proximoventrally and extended directly toward the Trl cell (open arrowhead). This forms the d segment (see
text). The Til growth cones are spreading over the Trl neuron, which has begun axonogenesis in the ventral direction, along the distal segment
boundary of the coxa. A cell (asterisk) labeled with anti-HRP lies near the point of reorientation, but is not precisely at it. A lamellum (star) is
extended lateral to the Til axons where they have crossed the tibia-femur segment boundary (see Caudy and Bentley, 1986). The Til axons also
make a distinct reorientation (arrow at right) toward the Fe1 cell after initially growing proximally along the a segment. Thus, in this leg the
characteristic a, b, c, and d segments all are exhibited in the axon morphology of a growth cone that has extended from the Til cell bodies to the
Trl neurons. Calibration bar, 50 Frn.
(Fig. 30, extendsdirectly to and along them, and endsat their
proximal edge.
Segmentg: This segmentbeginsat the proximal edgeof the
Cxl neurons, extends straight proximally to the CNS without
further reorientations(Figs.4A, 7B), and endswith contact with
the CNS.
In summary, in normal embryos the Til axons are oriented
in a characteristic direction at each point along their route to
the CNS. Furthermore, since axon orientations reflect the orientation that the growth cone had as it grew, the growth cones
apparently werespecificallyoriented at all points alongthe path,
and thus apparently did not grow randomly at any point. Fur-
Caudy and Bentley
Figure 2. Til morphologies along the c, d, and e segments. A, The Til
axons have extended straight from their cell bodies to contact a faintly
labeled Fe 1 cell (solid arrowhead). The Ti 1 axons have their characteristic proximodorsal orientation along the c segment, to a point (arrow)
where they distinctly reorient and extend directly to the Trl cell site,
to form the d segment. A single, small branch also extends dorsally from
the reorientation point (arrow). Calibration bar, 50 pm. B, The Til
axons have extended straight from their cell bodies to a point from
which lateral processes have contacted a faintly labeled Fe1 cell (solid
arrowhead). From that point they extend along the c segment in the
direction of the dorsally located ml cell (carat), which has labeled with
anti-HRP in this leg (c$ Figs. 1C and 6C,l. The proximodorsal orientation is maintained to the point (arrow) where the Til axons distinctly
reorient and extend directly along the d segment to a faintly labeled Tr 1
cell (open arrowhead). The Trl cell lies on the distal segment boundary
of the coxa. After contacting that cell, the Til growth cones have reoriented circumferentially along the segment boundary (solid triangle),
to begin forming the e segment (see text). As in A, Til processes (several
filopodia) extend straight dorsally from the point of reorientation (urrow). In this leg, a second dorsally located cell, m2 (open triangle) has
labeled with anti-HRP. A broad lamellar process is extended ventrally
from the second dorsal cell. Calibration bar, 50 pm.
Vol. 6, No. 6, Jun. 1966
Figure 3. Growth cone morphologies along the e and f segments. A,
The Ti 1 axons extend proximally to the distal segment boundary of the
coxa (solid triangles), where their growth cones abruptly reorient and
extend ventrally along the segment boundary. Curved arrow, Cxl cells.
Calibration bar, 50 pm. B, The same leg at higher magnification. Til
growth cones extend along the segment boundary to form segment e
The Journal of Neuroscience
1785
Pioneer Growth Cones Steer to Different Guidance Cues
Figure 4. Til morphologies and anti-HRP labeled cells along the g segment. A, The Til growth cones (arrow) have nearly reached the boundary
of the CNS (white square). The Til axons are straight and unbranched along the g segment (i.e., the 60-70 pm region straight distal to the growth
cones), which lies between the Cxl cells (unlabeled in this leg, but see Figs. 7 and 9) and the CNS. Caret, Efferent processes; solid arrowhead, Fe1
cell. Calibration bar, 100 pm. B, The pair of Cxl guidepost cells (curved arrow) has begun axonogenesis before the Til growth cones have reached
them. In this leg, another proximal (p) cell (straight arrow) on the final path has labeled with anti-HRP. This p cell lies about 15 pm straight
proximal to the Cxl cells. The Til growth cones have branches extended along the c segment. A single, very thin branch (upward, narrow arrow)
is in contact with a labeled Tr 1 neuron (not marked). White sauare. Boundarv of CNS: caret. efferent fibers: solid arrowhead. Fel. Calibration bar.
100 Wm. C’, The Til axons have reached the CNS boundary (kite square). Two adjacent p cells (arrows) have labeled with anti-HRP. The more
proximal of these cells is within one cell diameter of the CNS. The more distal of these p cells lies about 20-30 pm proximal to the site where the
Cxl cells are located. Caret, Efferent fibers. Calibration bar, 100 pm. D, The Til axons have extended proximally to contact the Fe1 cell (solid
arrowhead) and the Trl cell (open arrowhead), and have then turned circumferentially
along the distal coxal segment boundary. The axons have
turned proximally from the segment boundary and extended to the Cxl cells (curved arrow). Two additional p cells (straight arrows) lie straight
Proximal to the Cxl cells. so that toaether these cells form a continuous chain along the final path that reaches nearly to the CNS (white square).
&et, Efferent fibers. Calibration bay, 100 pm.
since single, distinct reorientations typically initiate
the characteristic segments, it would appear that at these points
the growth cones made a discrete switch in the guidance cue
that they were following. We will therefore consider these characteristic segments of the path with regard to known or possible
guidance cues.
thermore,
Growth cone interactions with
guidepost neurons
The best characterized guidance cues for the Til growth cones
are identified guidepost neurons. In normal embryos the Til
t
(horizontal arrow). Three filopodia (downward arrow) are in direct contact with the Cxl neurons (curved arrow). Calibration bar, 50 pm. C,
The Ti 1 axons extend proximally to the Fe 1 neuron (solid arrowhead)
and then to the Trl neuron (open arrowhead). At the Trl neuron, they
reorient circumferentially along the distal segment boundary of the coxa
(solid triangles) to form segment e. They then make a single, distinct
reorientation toward the Cx 1 neurons (curvedarrow) and extend directly
to them to form the characteristic f segment (downward arrow). The
Til growth cones are at the Cxl neurons and are wrapping branches
and filopodia around one of them. Calibration bar, 50 pm.
axons make direct contact with these cells and selectively dyecouple with them (Bentley and Keshishian, 1982a, b, Keshishian
and Bentley, 1983a; Taghert et al., 1982). Deletion of the Cxl
guidepost cells has shown them to be a necessary cue for guidance along the normal path (Bentley and Caudy, 1983a, b).
Axon reorientations toward guidepost neurons
Three of the characteristic Til axon segments-b, d, andf-are
each oriented toward, make direct contact with, and then extend
along the Fe1 , Trl, and Cx 1 guidepost neurons, respectively.
The point at which the Til axons leave a guidepost neuron,
after extending along it, typically is the precise point where the
axons make a distinct reorientation onto the next characteristic
axon segment (Figs. 1, C, D, 5D). Thus, that point of reorientation clearly defines the termination of the previous axon segment.
Growth cone branches andjlopodia
wrap guidepost
selectively
neurons
Growth cones do not simply extend along guidepost neurons
that they have contacted. Frequently growth cones also profusely
wrap labeled guidepost neurons with filopodia (Fig. 1, C, D, 3C)
1786
Caudy and Bentley
Vol. 6, No. 6, Jun. 1986
Figure 5. Selective wrapping of unlabeled Fe1 and Trl guidepost cells by Til branches and filopodia. A, The Til axons have extended proximally
from their cell bodies and have extensively wrapped an unlabeled Fe1 cell (solid arrowhead) with branches and filopodia. Two branches have
extended beyond the Fe1 cell: one (upward arrow) directly toward the site where the Trl cell arises, and one (downward arrow) proximodorsally
along the characteristic c segment toward the region where the ml cell occasionally labels with anti-HRP (see Figs. 1C and 6, C, D). A prominent,
triangular-shaped lamellum (star) is extended laterally from the Ti 1 axons, where they have crossed the tibia-femur segment boundary (see Gaudy
and Bentley, 1986). Curved arrow, Cxl cells. Calibration bar, 50 pm. B, The Til axons have extended proximally and have wrapped an unlabeled
Fe 1 cell (solid arrowhead) with branches and filopodia. At least one Til axon (arrow) has extended further proximally and is wrapping an unlabeled
Trl cell (open arrowhead) with branches and filopodia. Curved arrow, Cxl cells. Calibration bar, 50 pm. C, The Til axons have wrapped an
unlabeled Fe 1 cell (solid arrowhead) with branches and filopodia, which extend around the complete circumference of the cell on the inner epithelial
surface. Two Til processes have extended beyond the Fe1 cell: a prominent branch (upward arrow) is extended directly to the site where the Trl
cell arises, and several filopodia (downward arrow) are extended proximodorsally
toward the region where the ml cells are typically located (see
Figs. lC, 6, B, C, D). Calibration bar, 25 Nrn. D, The Til axons (arrow) have extended proximally and have contacted an unlabeled Trl cell (open
arrowhead). Branches and filopodia have wrapped in both directions around the cell on the inner epithelial surface (the plane of focus), and additional
filopodia are curving over its surface. The Tr 1 cell lies on the distal boundary of the coxal segment (solid triangle) and a branch is extended ventrally
along the segment boundary. Calibration bar, 25 pm.
and occasionally spread lamellae over them (Fig. 1D). Branches
also are observed to wrap over labeled guidepost neurons and
to extend around the circumference of the cell where it contacts
the epithelial substrate (Figs. 3, 4B).
The characteristic wrapping of filopodia and branches around
neurons frequently reveals the presence and location of unlabeled guidepost neurons (Fig. 5). Filopodia selectively wrap single unlabeled cells at the Fe1 site (Figs. 5, A-C, 6A), the Trl
cell site (Fig. 5, B, 0) and at the site of the Cxl cells (not shown).
Branches extended around the circumference of unlabeled cells
also are observed at the Fe1 site (Figs. 2A, 5, A-C, 60, the Trl
cell site (Fig. 5, B, D), and the site of the Cxl cells (not shown).
The selective wrapping of Til growth cone filopodia and
branches around guidepost neurons suggeststhat growth cones
specifically respond to contact with these cells. Furthermore,
these characteristics allow Til morphologies to be used as indicators of the presence of unlabeled guidepost cells on the path.
Til morphologies and putative cuesfor the
other characteristic segments
In any given embryo, Til growth cones extending along the
characteristic segments that are not oriented toward guidepost
The Journal of Neuroscience
Pioneer Growth Cones Steer to Different Guidance Cues
1787
C
Figure 6. Ti 1 growth cone orientation along the c segment. A, The Ti 1 axons have contacted an unlabeled Fe 1 cell (solid arrowhead) and wrapped
it with filopodia. At the Fe1 cell site, the pair of Ti 1 growth cones has separated (long arrows), and each has independently oriented proximodorsally
along a cluster of leading filopodia. The Trl cell (open arrowhead) has begun to label with anti-HRP and lies straight proximal to the Fe1 cell, but
the Til growth cones are oriented toward a region dorsal to the Trl cell, along the characteristic c segment. Curved arrow, Cxl cells. Calibration
bar, 50 Mm. B, The Ti 1 axons have extended proximally to the Trl cell site (open arrowhead) and are wrapping branches around an unlabeled cell
at that site. A prominent branch (downward arrow) also has extended proximodorsally
from the Fe1 cell site and has divided into 3 smaller branches
that converge at the site of the ml cell (caret; CJ Fig. lc). Curved arrow, Cxl cells. Calibration bar, 50 pm. C, The Til axons have extended
proximally from their cell bodies. Several filopodia (arrows) are extended parallel to one another from different sites on the Til growth cone. They
are extended proximodorsally
toward the ml cell (carat), which is apposed to the inner epithelial surface. Anti-HRP-labeled
particles extend radially
from the ml cell and may reflect lamellar processes (cf: Fig. lC’), which the Til filopodia may contact. Calibration bar, 50 pm. D, In another leg,
the ml cell (caret) has again labeled with anti-HRP, and several filopodia (urrows) extended proximodorsally
from the Til axons have made direct
contact with it. At least 2 of the filopodia divide in 2 very near the point where they contact the ml cell. Calibration bar, 25 pm.
neurons are as highly oriented as growth cones growing toward
guidepost cells. Growth along each of the “nonguidepost” segments is initiated
by a single distinct
orientation
(segment
a) or
reorientation, as though each resulted from a separate guidance
cue. However, growth cones on these segments do not appear
to be oriented by a specific, single cell at the termination point.
Segmenta: proximal outgrowthfrom the
Til cell bodies
Ti 1 growth cones and axons do not exhibit filopodial wrapping
with labeled or unlabeled cells at the termination point of the
a segment (Fig. 1, A, B). Also, the a segment varies in length
by a greater amount than do the b and d segments (Table 1).
These observations suggest that the a segment is not oriented
by a specific cell or feature at a fixed distance from the Til cell
bodies.
The a segment generally is oriented along the geometrical axis
of the tibia, in which the Ti 1 neurons arise (schematized in Fig.
10). The pair of Til cell bodies are usually aligned along this
axis (Figs. 1, 5, A, B, and 64, so that their axons emerge from
the proximal poles of the cell bodies and extend straight proximally until they reorient toward the Fe1 neuron (Fig. 1, A, B,
D). Therefore, proximal axonogenesis along the a segment may
be due to internal polarity of the Til cell bodies. Alternatively
(or in addition), there also is evidence that proximal guidance
of the Ti 1 growth cones in this region may result from proximal
increases in the affinity of the substrate epithelium for growth
cones (Caudy and Bentley, 1986).
1788
Caudy and Bentley
Vol. 6, No. 6, Jun. 1986
some other cell types also label (Jan and Jan, 1982; Shankland
and Bentley, 1983). However, the ml cell does not appear to
be a neuron. Unlike known neurons, it only occasionally labels
with anti-HRP. Furthermore, even when it does label, it apparently does so only transiently, as labeled ml cells are not
observed at later stages.Its bipolar morphology and axial alignment (Figs. 2B, 60), its broad lamellae apposed to the inner
epithelial surface (Fig. 1C), and its characteristic dorsal location
suggest that it is one of a set of identified muscle pioneer cells
in this region (Ball et al., 1985; Ho et al., 1983; Robert Ho,
personal communication).
Filopodia regularly appear to be oriented specifically toward
the ml cell (Figs. lC, 6, C, D). On occasion, multiple filopodia
directly contact this cell (Fig. 60). Furthermore, after the Til
growth cones have reoriented and extended to the Trl cell,
clusters of filopodia are regularly observed extending from the
reorientation site toward the ml cell (Figs. 5, A, C, 8).
Other factors may also contribute to guidance along the c
segment. The c segment does not always orient directly toward
the ml cell (Fig. 8). Another cell, m2, occasionally labels in the
same dorsal region as ml, and may contribute to growth cone
guidance along segment c. The segment boundary between the
femur and trochanter also develops near the termination point
of the c segment (Caudy and Bentley, 1986) and may also contribute to Til morphologies there. Furthermore, the distinct
proximodorsal reorientation that initiates growth along the c
segment occurs before the growth cones appear to be in filopodial contact with the ml cell or other dorsally located cells
(Fig. 1B). Thus, there may be other, unknown cues involved in
the initiation of growth along the c segment. However, the ml
cell frequently does appear to influence the Til growth cone in
the terminal region of the c segment (Figs. 60,8). We conclude
that ml contributes to branch extension and growth cone morphology on segment c, but that other factors may be involved.
.
Cxl growth cone and axon morphologiesin the g segment
region. A, In some legs,the Cxl neurons (curvedarrow) pioneer the
same pathway to the CNS (white square) that forms the g segmentof
the Til axons. When the Cxl growth conesextend proximally along
this path, their growth conesare highly oriented straight proximally,
evento the degreethat filopodia tend to lie in very tight clusters(arrow).
Calibration bar, 50 pm. B, At a later stage,when the Cxl (curvedarrow)
growth coneshave reachedthe CNS (white square), their axons(arrow)
are straight.The Cxl axonshaveno lateral branchesand very few lateral
filopodia. Caret, Efferent fibers. Calibration bar, 50 pm.
Figure 7.
Segmentc: proximodorsal reorientation
from the Fe1 neuron
The c segment typically is initiated by a single distinct reorientation, suggesting a discrete switch in guidance cue. However,
growth cones and axons on the c segment do not exhibit any
specific interactions with labeled or unlabeled cells at the termination point of the segment. Furthermore, separated Til
growth cones (Fig. 6A) and filopodia (Fig. 6C) can orient proximodorsally from different points and extend roughly parallel.
They are not oriented toward a single point in the region where
subsequent reorientations typically occur. In addition, branches
regularly extend proximodorsally beyond the point of reorientation toward Trl (Fig. 8). Also, branches occasionally (although
very rarely) extend beyond the typical reorientation point and
continue proximodorsally to the dorsal edge of the leg, where
they appear to converge on an unlabeled cell (Fig. 6B).
In some legs, a specific cell, ml, transiently labels with antiHRP in this same region at the dorsal edge of the leg (Figs. 1C,
6, C’, D). Anti-HRP antibodies typically are specific for a cellsurface molecule on insect neurons (Jan and Jan, 1982), although
Segmente: ventral reorientation
from the Trl neuron
On leaving the Trl guidepost neuron, the Til growth cones
(Figs. 2B, 5D) and axons (Fig. 3; see also Bentley and Caudy,
1983b; and Caudy and Bentley, 1986) distinctly reorient circumferentially and extend ventrally to form the characteristic
e segment. The guidance cue for this segment of growth has been
identified as a limb segment boundary (Bentley and Caudy,
1983b), in particular, the distal segment boundary of the coxa
(Caudy and Bentley, 1986). This segment boundary differentiates within the epithelial substrate. Once on this continuous
circumferential cue, growth cones apparently extend circumferentially along it until filopodial contact with the proximally
located Cx 1 neurons distinctly reorients them proximally (Fig.
3; Bentley and Caudy, 1983b).
Segmentg: proximal reorientation
from the Cxl neurons
On leaving the Cx 1 guidepost neurons, the Ti 1axons and growth
cones orient proximally, along the g segment. The g segment
characteristically is very straight, without lateral branches of
either Til growth cones or axons (Fig. 4A).
Unlabeled axons from the Cxl cells could provide a continuous cue for Til growth cones on the g segment, since they
sometimes pioneer the path to the CNS (Fig. 7). However, when
the Cxl cells pioneer the pathway, they too are very highly
oriented and do not branch (Fig. 7). The lack of lateral processes
extends even to the lack of lateral filopodia (Fig. 7). One possible
cue for guiding the Ti 1 and Cx 1 growth cones along the g segment is a set of proximal cells, the “p” cells, which appear to
form a nearly continuous chain there (Fig. 4, B, C’, D). These
cells sporadically label with anti-HRP and may be present in
all legs. Growth cones occasionally exhibit a low degree of fil-
The Journal of Neuroscience
Pioneer Growth Cones Steer to Different Guidance Cues
1789
opodial wrapping of at least the most distal of these cells (Fig.
4B). However, growth cones (Fig. 7A) and axons (Figs. 4A, 7B)
on the g segment typically exhibit no branch or filopodial wrapping of cells there. We conclude that there is a strong extrinsic
factor that orients growth cones on the g segment and that the
p cells may sometimes contribute to guidance there.
Normal variations in the axon path
Seven characteristic axon segments in the overall Ti 1 path have
been described. Different axon segments exhibit different degrees of variability in form and expression. The g segment always is present, and always extends straight proximally from
the Cxl neurons. The f segment also is always present, and
always extends to the Cxl neurons. However, the point at which
the f segment begins can range from the point at the ventral
edge of the Trl cells to any point more ventral along the distal
segment boundary of the coxa, up to the point on the boundary
that is straight distal to the Cxl cells.
The e segment is totally absent in pathways in which the Til
axons extend directly from the Trl neurons to the Cx 1 neurons.
When it is present, it always is oriented straight ventrally along
the segment boundary and ranges in extension from the Trl
neurons to the point straight distal to the Cx 1 neurons (Bentley
and Caudy, 1983b; Caudy and Bentley, 1986).
The axon segments between the Ti 1 and Fe1 neurons can
vary both in the exact direction of orientation, and in whether
they are present. The orientation of the a segment can range
over about 30” of arc, between the line extending straight proximal along the geometric axis of the limb, and the line extending
straight to the Fe 1 neuron (schematized in Fig. 10, upper right).
Axons occasionally extend straight from the Ti 1 neurons to the
Fe1 guidepost neuron (Figs. 2A, SB). In such cases it is not
possible to identify this straight segment as an a segment or a
b segment, nor to determine if or where a switch in guidance
cue was made. In other cases,the b segment is definitely missing
(Fig. 2B; see also Caudy and Bentley, 1986, fig. 2, A, B), and
the Ti 1 axons do not directly contact a labeled Fe 1 cell but pass
dorsal to it.
Similarly, the axon segments between the Fe1 and Trl neurons can vary, both in whether they are present and in their
exact orientation. Axons occasionally extend straight between
these neurons, so that there is no apparent c segment (Fig. 5,
C, D). However, even where there is no c segment, filopodial
extension often is biased dorsally (Figs. 5C, 6C). In rare cases,
axons can orient ventrally before reaching the distal coxa segment boundary and thus miss Tr 1 (Fig. 4C; seealso Caudy and
Bentley, 1986); in these cases, both the c and d segments are
missing.
The range of variation of axon pathways in normal embryos
is schematized at the upper right of Figure 10. Other extreme
variations occur in specially selected clutches of embryos where
the Ti 1 growth cones extend proximally before the Fe 1 and Tr 1
guidepost neurons have differentiated (Caudy and Bentley, 1986).
In such embryos, straight axon pathways from the Ti 1 neurons
to the distal boundary of the coxa are regularly observed, and
pathways frequently do not contact the Fe1 and Trl neurons
once they differentiate (Caudy and Bentley, 1986, fig. 8F). Also,
in these embryos growth cones in this region exhibit profuse
branching and are not highly oriented (Caudy and Bentley, 1986).
Orientations and reorientations of
Til growth cones
Growth cone morphologies confirm the above observations for
axon segment orientations. Til growth cones in normal embryos
have very few, if any, branches (Fig. 1). As a result, the orientation of the growth cone as a whole often can be clearly defined
by the orientation of the axon or major branch trailing immediately behind the leading edge of the growth cone. Growth
f
I
VENTRAL
d-AXONS
~-BRANCHES,
FILOPODIAL
(n =I51
CLUSTERS
Figure 8.
CharacteristicTi 1branchesand filopodial clustersat the site
of reorientation toward the Trl neurons. Upper diagram, Schematic
representation of a Til neuron with axon segments along the a and b
segments that has extended along the c segment to the point (caret) of
reorientation toward the one Trl neuron. Branches are drawn in each
of 3 directions characteristic of branch and filopodial cluster extension
from that point: (1) directly toward the Trl neuron; (2) proximodorsally
toward the ml cell; and (3) straight dorsally from the reorientation point
(possibly toward the m2 cell, or alongthe prospectivefemur-trochanter
segmentboundary).Lower diagram, The reorientation site (carat) lies
at the centerof a polar plot of the directionsof Ti 1 axons (solid triangles)
and branchesand filopodial clusters(open triangles) observed to extend
from that point in 15 representative examples. Axons extendedonly to
the Trl cell, never in the other directions. Branches and tilopodial
clusters characteristically are oriented in the direction of the ml cell,
and also straight dorsally. No branches or clusters are observed in other
directions, nor do cells label in other directions. The inwardly oriented
solid triangles, lower right, indicate the directions from which the Til
axons approach the reorientation site along the c segment. Note that
these directions are frequently not oriented directly toward the ml cell,
although they may be in individual cases (e.g., Figs. 1C, 48).
Caudy and Bentley
1790
Fe1
Ti
- 1
u
rl = 33.8
AI 8.5ym
t-2 = 31.7
2 6.lym
(n=lO)
(n=ll)
Figure 9. Averagedistance at which Ti 1 axons reorient to the Fe I and
Tr 1 neurons.The drawing schematizesa growth conethat hasextended
along the characteristica, b, and c segmentsand is at the reorientation
site that initiates the d segmentthat extendsto the Trl neuron.The rl
and r2 distancesare the averagedistancesmeasuredfrom the reorientation sitesthat initiate the b and d segments,respectively,to the point
of contact with the Fe1 and Trl neurons.
cones at different locations along the path generally have the
same orientation as do nascent axon segments observed at the
same location in legs fixed at later stages. Thus, highly stereotyped orientations characterize both axons and growth cones at
all points along the Til pathway.
Disposition of growth conebranches
Growth cone branches do not appear to be extended randomly
in normal embryos. They characteristically extend around adjacent guidepost neurons or are oriented along the segments of
the ultimate Til path (e.g., Figs. 1C, 4B, 54, 6, A, B). Usually
only single branches are oriented toward guidepost neurons (Figs.
1C’, St); multiple branches are sometimes observed oriented to
other cues (Figs. 4B, 6B). Branches that are extended lateral to
the axon path are regularly observed at only 3 locations: (1)
circumferentially at the distal segment boundary of the femur
(Caudy and Bentley, 1986); (2) dorsally along the distal segment
boundary of the coxa (Caudy and Bentley, 1986); and (3) at the
site of reorientation that terminates the c segment (Figs. 2,4, 8).
Branch and filopodial extension at segment boundaries is described elsewhere (Caudy and Bentley, 1986). We conclude that
in normal embryos: growth cone branches are extended only
toward or along specific extrinsic guidance cues.
Dispositionof growth conefilopodia
Filopodia of Til growth cones fixed in situ are not disposed in
a symmetrical fan at the leading edge of the growth cone, as is
often seen in growth cone filopodia in culture (Bray, 1982; Letourneau, 1979). Rather, filopodia at the leading edge of Til
growth cones are predominantly (but not exclusively) oriented
in the direction of extrinsic cues (Figs. 1, B, C, 2B, 3B, 5, A, C,
6, A, C, D, 7A, 8). Single filopodia are regularly extended toward
distant guidepost neurons in regions where there are no adjacent
filopodia of comparable length (Fig. 1C). In addition, clusters
of filopodia (defined in Materials and Methods) are predominantly oriented to specific cues, along the c segment (Figs. 1, B,
C, 5A, 6, A, C, 8), along the e segment (Pig. 2B), and along the
g segment (Fig. 7A). We conclude that the prominent filopodia
observed to contact guidepost neurons and the distinct filopodial
clusters oriented toward other cues reflect growth cone response
to specific extrinsic cues.
Growth conereorientationprocesssuggested
by stereotypedTil morphologies
The distinct reorientations of axon segments toward guidepost
neurons suggestthat growth cones also distinctly reorient toward
Vol. 6, No. 6, Jun. 1986
them. Although the dynamic process of growth cone reorientation cannot be observed in fixed tissue, the stereotyped morphologies of Ti 1 growth cones suggest a particular sequence of
growth cone response and reorientation to guidepost neurons.
Growth cones are observed that exhibit different degrees of
orientation toward, and contact with, the Tr 1 guidepost neurons:
(1) growth cones oriented along the c segment, which have no
processes oriented toward the Trl neuron, although that cell
has begun to label with anti-HRP and is therefore definitely
present (Fig. 6A); (2) growth cones oriented along the c segment,
which have filopodia in direct contact with the Trl neuron but
have no branches oriented toward that cell (not shown); (3)
growth cones that have branches oriented directly toward the
Trl neuron along a filopodium already in contact with it, although other growth cone branches and filopodial clusters also
are present (Fig. 1C, 4B); (4) growth cones that have completely
reoriented toward the Trl neuron, in that no branches are extended in other directions (Fig. 1D).
Similar observations have been made for growth cones that
exhibit different degreesof orientation toward, and contact with,
the Cx 1 guidepost neurons: (1) growth cones that have oriented
ventrally along the e segment and have no processes oriented
toward the Cxl neurons (Figs. 2B, 5D); (2) growth cones that
have lateral filopodia in contact with the Cx 1 neurons and also
have branches extended further ventrally along the segment
boundary cue for the e segment (Fig. 3, A, B); (3) growth cones
that have a branch oriented toward the Cxl neurons along filopodia in contact with them, although a branch is still further
extended along the segment boundary (not shown here-see
Bentley and Caudy, 1983b, fig. 40); (4) growth cones that have
completed reorientation toward the Cxl neurons, with no
branches oriented further along the segment boundary (Fig. 3C).
The above observations suggest that growth cone branches
are guided directly to guidepost neurons by extending along
filopodia that have already made contact with those neurons.
Furthermore, reorientations toward guidepost neurons are not
observed in situations where the growth cones could not be in
contact with an unlabeled guidepost cell. Together these observations also suggest that growth cones reorient by withdrawing
branches extended in other directions and continuing to extend
along branches oriented toward, and in filopodial contact with,
guidepost neurons. Thus, the point from which a branch is
extended toward a guidepost neuron apparently becomes the
point of distinct, single reorientation observed between 2 adjacent axon segments at a later stage.
Axon reorientationstoward guidepostneurons
occur at characteristicdistances
To determine whether growth cones generally could contact
guidepost neurons with filopodia from the points where they
reorient, we have measured the distance at which axons reorient
toward the Fe1 and Trl guidepost neurons. The averages of
these distance are the rl and r2 distances depicted in Figure 9.
On the average, Ti 1 growth cones reorient toward both the Fe 1
and Trl neurons at a distance of about 30-35 pm. This distance
often is spanned by Til filopodia (Figs. lC, 3B). Thus, these
results support the hypothesis that the Ti 1 growth cones contact
distant guidepost neurons with filopodia, and then reorient directly toward them by extending branches along those filopodia.
Discussion
Til morphologies reveal a sequenceof distinct
reorientations toward difirent cues
The characteristic orientations of nascent axon segments are the
clearest indication of guidance by a sequence of different cues
(Figs. 1, 2, 3, 4A, 7B, schematized in Fig. 10). However, the
same orientations also are evident in the leading ends of growth
The Journal
of Neuroscience
Pioneer Growth Cones Steer to Different Guidance Cues
Til
D
/-------x
/
Pro/
\
V
i ‘*Dist
Figure IO. Summary diagram of the features of the leg and Til morphologies observed between the 33 and 35% stages. The pair of sibling Til
neurons have cell bodies that generally are aligned along the geometrical axis of the limb in the tibia (dashed line). The Til growth cones emerge
from the proximal poles of these cells and generally extend straight proximally along the limb axis, thus forming the characteristic a segment. After
about 50 km of proximal growth, the Ti 1 growth cones (in non-PPD clutches; cf: Caudy and Bentley, 1986) typically make a distinct reorientation
in the proximoventral
direction and extend directly toward and make contact with the Fe1 guidepost neuron at the midpoint of the femur (Fe),
thus establishing the b segment. From the Fe1 cell, the Til growth cones typically orient proximodistally
toward the region where the ml cell labels
occasionally, thus forming the c axon segment. After extending in this direction, the Til growth cones again distinctly reorient in the proximoventral
direction and extend directly toward and make contact with the Trl guidepostneuron. This establishesthe d axon segment.After contactingthe
Tr 1 neuron, the Ti 1 growth cones and axons reorient circumferentially
and extend along the distal segment boundary of the coxa. The e segment
results from ventral growth along the segment boundary. Thefaxon segment then is initiated by a distinct reorientation toward and direct extension
to the Cxl guidepost neurons. After leaving the Cxl cells, the Til growth cones and axons extend proximally to the CNS, forming the g axon
segment. A chain of cells, the p cells, sporadically and transiently labels with anti-HRP along the final path. The stippledregionindicates the regions/
orientations in which Ti 1 branches normally occur. These include regions along the tibia-femur segment and trochanter-coxa
segment boundaries
(seeGaudy and Bentley, 1986),as well as characteristicbranchesextendedtoward these ml and also straight dorsally from the reorientation site
between the c and d axon segments. lateral branches are not observed on the g segment. Til growth conesand axons sometimes extend directly
between neuronson the pathway, and the Til axon morphology ranges between direct growth between these cells and the indirect, zigzag routes.
This normal range of axon locations is indicated by the shaded regionsadjacent to the axon segments depicted in the upper-rightinset.
along the path (Figs. 1, 2, 3C, 5A, 6, A,
B, 7A, and 8). Furthermore, these same orientations are observed for individual branches of growth cones, which appear
to extend exclusively to specific cues (Figs. 1C, M, 8). Finally,
single lilopodia (Figs. 1C, 4B) and filopodial clusters (Figs. 1,
B, C, 2B, 7A, 8) also tend to orient in the direction of extrinsic
cues.
cones, as they extend
Til
growth
cones that grow
through
the same region
of the
leg before specific cellular cues have differentiated do not exhibit
the same characteristic orientation of growth cone processes
(Caudy and Bentley, 1986). Thus, these specific orientations of
Til
processes
are not due to intrinsic
programs
of neuronal
morphogenesis but, rather, to response to extrinsic cues. (This
also suggeststhat the distinct reorientations that have been analyzed here do not result from morphological deformation during
tissue mounting.)
Guidance cues for the Til
pioneer growth cones
The known and putative cues for the 7 characteristic axon segments are listed in Table 1 (see also Table 2). Cell-ablation
experiments have shown the Cxl neurons to be necessary and
sufficient cues for guidance along the f segment (Bentley and
Caudy, 1983a, b)- Ti 1 growth cones and axons typically exhibit
the same specific interactions with the Fe1 and Trl neurons as
they do with the Cxl guidepost neurons: (1) selective dye-cou-
pling
(Bentley
and Keshishian,
1982a,
b; Taghert
et al., 1982);
(2) selective branch and filopodial wrapping; and (3) characteristic reorientations
toward, direct extension to, and further extension along those neurons. Thus, Ti 1 growth cones specifically
interact with, and are guided by, each of the three guidepost
neurons on the path.
Our results differ from those of Berlot and Goodman (1984),
who concluded that the Fe1 neuron does not differentiate in
time to influence
the orientation
or guidance
of the Til
growth
cones. In contrast, we observe characteristic wrapping of branches and filopodia around the Fe 1 neuron and, more importantly,
the distinct reorientation of the Til growth cones and axons
toward and along Fe 1. We conclude that in many embryos, the
Fe1 (and Trl) neurons differentiate early enough to serve as
guidepost neurons for the Til growth cones and that in such
embryos, the Ti 1 growth cones both respond to, and are guided
by, those neurons (see also Bentley and Keshishian, 1982a, b;
Keshishian and Bentley, 1983a).
There are cases, however, in which proximal axonogenesis
can begin before filopodial
contact is established
with the Fe1
neuron (Bentley and Caudy, 1983b, Berlot and Goodman, 1984).
In such cases, proximal outgrowth is along the characteristic a
segment and is oriented along the limb axis. Two possible guidance mechanisms for such proximal outgrowth have been suggested. The axial alignment of the Til cell bodies suggeststhat
there may be an internal polarity to the Til neurons such that
1792
Caudy and Bentley
growth cones emerge from their proximal poles and continue
proximally (Bentley and Caudy, 1983b). Alternatively, there
may be an extrinsic “limb axis cue,” such as an adhesion gradient (Bentley and Caudy, 1983b; Berlot and Goodman, 1984).
Evidence that the affinity of the epithelial substrate for growth
cones increases proximally in this region has been presented
elsewhere (Caudy and Bentley, 1986). Similar observations of
directed outgrowth without guidepost cell contact have been
made for pioneer axons in the wings of Drosophila (Blair and
Palka, 1985; Blair et al., 1985; Schubiger and Palka, 1985).
Til growth cones respond to cues of
diferent cell types
The Til growth cones respond to 3 different types of cells as
they grow. The various guidepost cells are immature neurons.
The guidance cue at the distal coxal segment boundary probably
derives from (non-neuronal) epithelial cells, which are of ectodermal origin (Bentley and Caudy, 1983b; Caudy and Bentley,
1986). Finally, the ml cell to which the growth cones respond
as they extend along the c segment is probably a muscle pioneer
(Ball et al., 1985; Robert Ho, personal communication), and
therefore of mesodermal origin.
Hierarchy of afinities
The Ti 1 pioneer growth cones exhibit a hierarchy of affinity for
their various guidance cues. The Fel, Trl , and Cxl guidepost
neurons are the dominant cues in that once filopodial contact
is established with a guidepost neuron, growth toward that neuron is completed before the Til growth cones reorient toward
another cue.
Guidance along the c segment may be provided by the ml
cell, and perhaps by additional cells in the dorsal region of the
leg. Whether the ml ceil is the cue for the initial reorientation
along the c segment is uncertain. However, the Ti 1 growth cones
definitely respond to it (Fig. 60): They regularly extend branches
and filopodial clusters dorsally well beyond the point at which
the growth cones reorient to the Trl guidepost neuron (Fig. 8)
and, in very rare cases, can continue to the dorsal edge of the
leg if they do not reorient to Trl (Fig. 6B). Despite these specific
responses, the Til growth cones normally reorient toward the
Trl cell before they have completed growth to the ml cell and
other cells in the dorsal region (Fig. 1, C, D). It therefore seems
probable that the Trl cell is a higher affinity cue than the ml
cell and other putative cues for the c segment.
A similar inference can be made of the relative affinities of
the distal coxal segment boundary and the Cxl guidepost neurons. The segment boundary can provide a high-affinity cue for
pioneer growth cones. Growth cones on the segment boundary
can follow a very straight circumferential path along it for as
much as 180” of arc around the circumference of the leg (Bentley
and Caudy, 1983b). However, Til axons always are observed
to reorient away from the segment boundary and to extend
directly to the Cx 1 neurons, although branches here also extend
beyond the reorientation point, and growth on the segment
boundary clearly could have continued (Caudy and Bentley,
unpublished observations). Thus, the Cx 1 neurons appear to be
a higher affinity cue than the segment boundary.
The reorientation from the a segment toward the Fe1 neuron
may or may not reflect a hierarchy of affinity. Orientation along
the a segment may be due to an intrinsic mechanism rather
than an extrinsic cue, in which case the reorientation toward
Fe 1 would not reflect a difference in affinities between extrinsic
cues.
The reorientations that occur as growth cones leave a guidepost neuron do not reflect orientation along a cue of higher
affinity, since no further guidance by that neuron is possible.
The subsequent continuation of growth apparently reflects the
absence of a specific “stop” signal for the Til growth cones,
Vol. 6, No. 6, Jun. 1986
rather than reorientation along a higher affinity cue. Thus, for
the Til growth cones, growth along a sequence of cues of different affinities does not require a continuous increase in affinity
of cues in order for navigation to continue.
Molecular aspects of the afinity hierarchy
The differences in affinity exhibited by the Ti 1 growth cones for
their various cues probably result from the differential expression of affinity molecules by them. There could be a difference
in the degree of expression of a single type of affinity molecule
on the different cues, or there could be different molecules for
some or each of the cues. Also, these affinities could be adhesive
in nature or could involve some active form of biochemical
signaling (see also Caudy and Bentley, 1986). The differential
expression of adhesion molecules on the cues might explain the
observed preferential disposition of filopodia toward specific
cues. If the cues are adhesive for filopodia, randomly extended
filopodia would be preferentially retained at cues, while other
filopodia are withdrawn (Bray, 1982, Letoumeau, 1983). However, it is also possible that the various cues are adhesive for
filopodia, but that some or all cues also interact actively. The
selective establishment of dye-coupling with the guidepost neurons (Bentley and Keshishian, 1982a, b; Keshishian and Bentley,
1983a; Taghert et al., 1982), indicates that these cells, at least,
may undergo more than simple adhesive interactions with pioneer growth cones.
Interestingly, cues of all 3 cell types label with anti-HRP,
which generally is specific for surface molecules on neurons (Jan
and Jan, 1982). Although the ml cell apparently is a mesodermal
cell, it still labels with anti-HRP. Similarly, the epithelial cells
at segment boundaries also tend to label with anti-HRP (Caudy
and Bentley, unpublished observations), and segment boundaries form high-affinity cues for the Ti 1 growth cones. Furthermore, the observed differences in affinity correlate with the degree of expression of the molecular binding site for anti-HRP.
The highest affinity cues are the guidepost neurons, which all
label strongly and consistently with anti-HRP by the 35% stage.
The lower affinity cues such as the ml cell and segment boundaries label weakly and inconsistently. Thus, the anti-HRP binding site might be a molecular cue for neuronal recognition, or
it might be a surface molecule whose expression is linked to the
expression of such a cue.
Characteristics
of growth cone steering
Growth cones steer by selective branch extension
The process of growth cone reorientation toward guidepost neurons was inferred from Ti 1 morphologies consistently observed
in fixed tissue (seeResults). Growth cone branches are observed
to extend along filopodia already in contact with guidepost neurons (Fig. 1C’). The growth cone apparently reorients by extending along that branch and by withdrawing branches extended in other directions.
This process is supported by the additional observation that
growth cone branches are not only directly extended to specific
cues, but are also exclusively extended to cues (Fig. 8). Thus,
the observed extension of a branch toward a guidepost neuron
probably in no case is due to random growth. This strengthens
the interpretation that branches are selectively extended along
filopodia in contact with guidepost neurons in response to that
contact.
An analogous sequence of growth cone reorientation toward
cues other than guidepost cells also can be inferred from stereotyped Ti 1 morphologies. For example, Ti 1 growth cones that
have reached the Fe1 cell typically extend clusters of filopodia
proximodorsally, along the c segment (Fig. 1B). At slightly later
stages, Ti 1 growth cone branches selectively extended along the
c segment are observed (Fig. 1C). At later stages, nascent axon
The Journal of Neuroscience
Pioneer
Growth
Cones
Steer to Different
segments are observed that are oriented along the c segment
(Fig. 1D).
Thus, both for reorientations toward guidepost neurons and
for reorientations to other cues, a distinct reorientation in the
nascent axon pathway is established by selectively extending a
branch along, or to, the cue for the next segment. This selective
extension of branches is in contrast to the extension of multiple,
variously oriented branches in clutches of embryos in which the
Til growth cones extend proximally before the Fe1 and Trl
neurons and other specific cues have differentiated (see discussion of “PPD clutches” in Caudy and Bentley, 1986).
Selective branch extension along
OrientedJilopodia
Filopodia in fixed tissue are not randomly oriented in advance
of growth cone branches, but themselves tend to be oriented to
the same cues. The disposition of filopodia that precede branches varies widely, apparently depending on the nature of the cue
being followed. Clusters of filopodia oriented directly ahead of
branches are regularly observed on the c segment (Figs. 1C, 6A),
on the e segment (Fig. 2B), and on the g segment (Fig. 7A). The
width of the cluster varies, being very narrow on the g segment
and relatively wide on the c and e segments.
Filopodial dispositions on the segment boundary cue for the
e segment vary greatly, perhaps due to variation in the degree
to which the boundary has differentiated at the time of growth
along it. Filopodial clusters are regularly observed there (Fig.
2B), but branches sometimes extend along it without preceding
filopodia (Fig. 5D). Growth cones also have been observed to
orient along the segment boundary following a single thin process,either a single filopodium or a very thin branch (seeBentley
and Caudy, 1983b, fig. 40; see also Shankland, 198 1, for a
discussion of similar growth cone morphologies).
In some embryos, the Til growth cones extend proximally
before the Fe1 and Trl guidepost neurons have differentiated,
apparently in response to a proximal increase in the affinity of
the epithelial substrate within segments (seediscussion of “PPD
clutches” in Caudy and Bentley, 1986). This cue appears to be
distributed over the epithelial substrate, and in PPD clutches
the Ti 1 filopodia appear to be more randomly oriented over the
epithelium. In general, the presence and disposition of filopodia
extended ahead of growth cone branches may depend on (1)
whether growth is along a continuous cue (such as the segment
boundary) or (2) is toward a distant cue (such as guidepost
neurons), (3) the relative affinity of the cue, and (4) the degree
to which the cue is localized.
Distinct, single reorientations imply
distinct switches in guidance cue
The single, distinct reorientations observed between nascent
axon segments suggestthat a discrete switch was made between
guidance cues. This interpretation is supported by the further
observation that at earlier stages,growth cone branches not only
are oriented specifically to cues, but are also oriented exclusively
to cues. At no stageare branches extended in intermediate (noncue) directions. Thus, at no stage or location does the direction
of an individual branch appear to reflect the sum influence of
multiple cues. This further suggeststhat a growth cone reorients
by making an abrupt “switch” or commitment to extension
along the branch oriented to the highest affinity cue, rather than
by a gradual reorientation as the growth cone progresses.
Direct filopodial contact guides branch
extension to guidepost neurons
The extension of growth cone branches toward distant guidepost
neurons apparently results from direct filopodial contact. The
consistent observation of growth cones with different degreesof
orientation toward, and contact with, the Tr 1 and Cx 1 guidepost
Guidance
Cues
1793
neurons suggestsa process by which Til growth cones reorient
to guidepost neurons (see Results). Direct filopodial contact
apparently triggers selective branch extension toward guidepost
neurons, as well as the eventual withdrawal of other branches
previously extended to other cues.
Several further observations support this hypothesis. Branches are not observed that are oriented in the direction ofguidepost
neurons but are clearly not, or could not be, in direct filopodial
contact with them. In addition, the distinct reorientations observed between nascent axon segments toward the Fe 1 and Tr 1
guidepost neurons both occur at a distance of 30-35 pm, which
corresponds to the filopodial reach of the growth cone (Bentley
and Caudy, 1983a, b, Bentley and Keshishian, 1982a, b; Keshishian and Bentley, 1983a; Taghert et al., 1982). The fact that
reorientations toward both of these neurons occur at about the
same distance suggests that the same guidance mechanism occurs in both cases. Furthermore, filopodial contact with the Cxl
neurons from growth cones on segment e also is frequently
observed at this distance (Fig. 3B).Therefore, direct filopodial
contact is a likely mechanism for selective branch extension,
and commitment to growth, to guidepost neurons.
Singlejlopodia can direct growth
cones to guidepost neurons
At all locations on the path, branches tend to be oriented to
specific cues (e.g., Fig. 8). Therefore, branches oriented in the
direction of a guidepost neuron are probably selectively extended to some cue in that direction rather than randomly extended. The observation of direct contact by a single filopodium
(Fig. 1C) suggestsboth that the neuron is the cue, and also that
single filopodial contact mediates selective branch extension. (It
also is possible that the initiation of branch extension requires
multiple filopodial contact, but, if so, the additional filopodia
must be withdrawn as the branch extends along a single filopodium.)
Til growth cones generally extend single branches toward
guidepost neurons. These single branches typically are extended
along single filopodia that are in direct contact with those neurons. Branches in fixed tissue are observed at several apparent
stages of extension along single filopodia to guidepost neurons:
(1) Filopodial contact is made before branch extension begins
(Fig. 3B); (2) a branch is extended partway along a single filopodium that is in direct contact with a guidepost neuron (Fig.
1C); (3) a branch is extended completely to the site of a guidepost
neuron (Fig. 5A); (4) the growth cone has reoriented and has
reached the guidepost neuron (Fig. 1D).
The progressive development of a branch extended along a
single filopodium also is suggested by the observation of extremely thin branches (Fig. 4B) extended to distant guidepost
neurons, as well as much thicker branches at presumably later
stages (Fig. 5A). Although branch extension to guidepost neurons may sometimes occur along multiple filopodia, single filopodial contact regularly, and perhaps typically, directs branch
extension to them.
Single filopodia can reorient growth
cones in the presence of
more prominent processes
Growth cones reorient toward guidepost neurons along single
filopodia, even though multiple filopodia, or even prominent
branches, still are preferentially oriented to the previous cue.
Although these more prominent processes are eventually withdrawn, and may be in the process of withdrawal at the time of
reorientation, they are present throughout the reorientation process, and sometimes for substantial periods afterward. For example, the multiple filopodia and/or branches that are selectively oriented along the c segment are not only still present
after the Til growth cones have reached the Trl neuron, but
1794
Caudy and Bentley
sometimes also after they have even reached the CNS (Fig. 4,
A, D). Thus, contact with a guidepost neuron by a single filopodium consistently will reorient a growth cone that is still
oriented to the previous cue.
Implications for models of growth cone steering
How does single filopodial contact with a guidepost neuron lead
to selective branch extension and commitment to growth along
that specific lilopodium while many other filopodia remain preferentially oriented to the previous cue?Two basic types of growth
cone steering models have been proposed. In their most extreme
form, “passive adhesion” models propose that contraction forces
generated within filopodia cause the retraction of nonadherent
filopodia and produce traction forces along adherent filopodia
that “pull the growth cone across the substrate” (Gunderson
and Park, 1984). In such models, each filopodium that is selectively retained by adhesive contacts would contribute to the
overall motive force exerted on the growth cone, and the final
direction of movement would be in the direction of the vector
sum of such forces. In contrast, “active” models involve biochemical response to contact with guidance cues, perhaps by
regulating cytoskeletal conformation (Cooper and Schliwa, 1985),
which might amplify relatively small differences between cues.
A problem with passive adhesion models is that differences
in adhesion may be too weak to steer growth cones, let alone
steer them along single filopodia. Growth cones in culture can
extend directionally on growth factor-conditioned substrates
where the spatial differences in the adhesivity of the substrate
are relatively small (about 20%; Gunderson and Park, 1984),
suggesting that other processesmay be necessary to steer growth
cones, or at least to amplify adhesion differences that are present
(Gunderson and Park, 1984).
Aligned microhlaments have been observed at the bases of
filopodia fixed in culture (Letoumeau, 1979). Two processes
have been suggested by which relatively small adhesion differences might lead to growth cones steering along such aligned
microfilaments: (1) Traction forces along lilopodia “may be directed to pull forward intracellular structures that interact with
actin filaments” (Letoumeau, 1979; seealso Bray, 1979), rather
than to pull the whole growth cone; or alternatively, (2) forces
generated along actin filaments within hlopodia “may be transmitted to microtubules and other structures [within the growth
cone] to determine their positions in the growth cone margin”
(Letoumeau, 1983). The directed orientation of microtubules
could then produce directed branch extension by an active process of adding monomers to the microtubule ends. In this way
an active process would amplify an initial response that was
based on relatively weak differences in adhesion.
However, the same alignment that mediates these two processescould also arise from an active response to a nonadhesive
cue. For example, a filopodium in contact with a guidepost
neuron might send a second messenger down its length to its
base, where it could, for example, cause a localized contraction
of the actomyosin known to be present there. Pioneer growth
cones selectively dye-couple with guidepost neurons (Bentley
and Keshishian, 1982a, b, Keshishian and Bentley, 1983a;
Taghert et al., 1982), and filopodia and neuronal cues also
undergo other active interactions (Bastiani and Goodman, 1984).
Thus, active responses to filopodial contact with guidepost neurons are a reasonable possibility.
Does differential adhesion mediate the steering response of
the Til pioneer growth cones? Recall that multiple filopodia
and/or branches are observed to be selectively oriented toward
the ml cell while the growth cone is in the process of reorientation to Trl and are often observed well after the completion
of that reorientation. According to strictly passive adhesion
models of steering, these processes should have been selectively
retained by adhesive contacts. As a result, they also should exert
Vol. 6, No. 6, Jun. 1986
traction forces on the growth cone that influence its direction
of growth. However, no such influence is observed, nor do Ti 1
growth cones steer in the direction of the greatest number of
selectively retained filopodia.
Thus, the Til growth cones do not appear to be steered by
the vector sum of lilopodial forces acting on the growth cone.
The growth cone neither advances nor is it pulled as a unit along
the epithelial substrate, since filopodial clusters and prominent
branches remain at their original locations for significant periods
of time after the growth cone reorients. Rather, the growth cone
can respond independently to multiple cues, although it consistently reorients to the cue of highest affinity, and eventually
withdraws branches previously extended to cues of lesser alhnity.
Finally, growth cones also appear to make a distinct switch
in the cue to which they are steering, and probably can do so
on the basis of contact by a single filopodium. Our results are
not consistent with a strictly passive adhesion model of growth
cone steering, but they are consistent with models of steering
by differential adhesivity of cues in which intracellular events
amplify adhesive interactions sufficiently for growth cones to
reorient along single filopodia. Alternatively, the results are also
consistent with purely active models of steering that do not
involve adhesion differences between cues, but, rather, are mediated solely by active guidance molecules.
References
Ball, E. E., R. K. Ho, and C. S. Goodman (1985) Muscle development
in the grasshopperembryo. I. Muscles, nerves,and apodemesin the
metathoracicleg. Dev. Biol. I1 I: 383-398.
Bastiani, M. J., and C. S. Goodman (1984) Neuronal growth cones:
Specific interactions mediated by filopodial insertionsand induction
of coated vesicles.Proc. NatL Acad. Sei. USA 81: 1849-1853.
Bate, C. M. (1976) Pioneerneuronsin an insect embryo. Nature 260:
54-56.
Bentley, D., and M. Caudy (1983a) Pioneer axons lose directed growth
after selective killing of guidepost cells. Nature 304: 62-65.
Bentley, D., and M. Gaudy (1983b) Navigational substratesfor peripheral pioneer growth cones: Limbxis
polarity cues, limb-segment
boundaries, and guidepost neurons. Cold Spring Harbor Symp. Quant.
Biol. 48: 573-585.
Bentley, D., and H. Keshishian (1982a) Pioneer neurons and pathways
in insect appendages.Trends Neurosci. 5: 364-367.
Bentley, D., and H. Keshishian
(1982b) Pathfinding by peripheral
pioneer neurons in grasshoppers. Science 218: 1082-1088.
Bentley, D., H. Keshishian, M. Shankland, and A. Toroian-Raymond
(1979) Quantitative stagingof embryonic developmentof the grasshopper,Schistocercanitens.J. Embryol. Exp. Morphol. 54: 47-74.
Berlot, J., and C. S. Goodman (1984) Guidanceof peripheralpioneer
neurons in the grasshopper: Adhesive hierarchy ofepithelial and neuronal surfaces. Science 223: 493-496.
Blair, S., and J. Palka (1985) Axon guidance in cultured wing discs
and disc fragmentsof Drosophila. Dev. Biol. 108: 4 11-4 19.
Blair, S., M. A. Murray, and J. Palka (1985) Axon guidance in cultured
epithelial fragments of the Drosophila wing. Nature 315: 406-408.
Bray, D. (1979) Mechanical tension produced by nerve cells in culture.
J. Cell Sci. 37: 391-410.
Bray, D. (1982) Filopodial contraction and growth cone guidance. In
Cell Behaviour:A Tribute to Michael Aberciombie, R. B&airs et al.,
eds., pp.
__ 299-3 17, Cambridae U. P., New York.
Caudy, M., and D. Bentley (1986) Pioneer growth cone morphologies
reveal proximal increasesin substrateaffinity within leg segmentsof
grasshopperembryos.J. Neurosci. 6: 364-379.
Cooper,M. S., and M. Schliwa (1985) Electrical and ionic control of
tissue cell locomotion in DC electric fields. J. Neurosci. Res. 13: 223244.
Gunderson, R. W., and K. H. C. Park (1984) The effects ofconditioned
media on spinal neurites:Substrate-associated
changesin neurite direction and adherence. Dev. Biol. 104: 18-27.
Ho, R. K., and C. S. Goodman
(1982) Peripheral pathways are pioneered by an array of central and peripheral neurones in grasshopper
embryos. Nature 297: 404-406.
The Journal
of Neuroscience
Pioneer Growth Cones Steer to Different Guidance Cues
Ho, R. K., E. E. Ball, and C. S. Goodman
(1983) Muscle pioneers:
Large mesodermal cells that erect a scaffold for develodnr! muscles
and-motoneurones
in grasshopper embryos. Nature 3Oi: 66.
Jan. L. Y.. and Y. N. Jan (1982) Antibodies to horseradish ueroxidase
as speciffic neuronal markers iu Drosophila and grasshopper embryos.
Proc. Natl. Acad. Sci. USA 79: 2700-2704.
Keshishian, H., and D. Bentley (1983a) Embryogenesis of peripheral
nerve pathways in grasshopper legs. I. The initial nerve pathway to
the CNS. Dev. Biol. 96: 89-102.
Keshishian, H., and D. Bentley (1983b) Embryogenesis of peripheral
nerve pathways in grasshopper legs. II. The major nerve routes. Dev.
Biol. 96: 103-l 15.
Keshishian, H., and D. Bentley (1983~) Embryogenesis of peripheral
nerve pathways in grasshopper legs. III. Development without pioneer
neurons. Dev. Biol. 96: 116-124.
Letoumeau, P. C. (1975a) Possible roles for cell-to-substratum
adhesion in neuronal morphogenesis. Dev. Biol. 44: 77-91.
Letoumeau, P. C. (1975b) Cell-to-substratum
adhesion and guidance
of axonal elongation. Dev. Biol. 44: 92-102.
1795
Letoumeau, P. C. (1979) Cell-substratum
adhesion of neurite growth
cones and its role in neurite elongation. Exp. Cell. Res. 124: 127138.
Letoumeau, P. C. (1983) Axonal growth and guidance. Trends Neurosci. 6: 451-455.
Schubiger, M., and J. Palka (1985) Genetic suppression of putative
guidepost cells: Effect on establishment of net% pathways-in Drosoohila winas. Dev. Biol. 108: 399-410.
Shankland, My (198 1) Embryonic development of a sensory afferent
projection in the grasshopper embryo. I. Growth of peripheral pioneer
axons within the central nervous system. J. Embryol. Exp. Morphol.
64: 169-185.
Shankland, M., and D. Bentley (1983) Sensory receptor differentiation
and axonal path6nding in the cercus of the grasshopper embryo. Dev.
Biol. 97: 468-482.
Taghert, P. H., M. J. Bastiani, R. K. Ho, and C. S. Goodman
(1982)
Guidance of pioneer growth cones: Filopodial contacts and coupling
revealed with an antibody to lucifer yellow. Dev. Biol. 94: 391-399.